1. Optimizing high-resolution troposphere estimates using PPP method and Benchmark data set
Václavovic P., Douša J.
Geodetic Observatory Pecný, RIGTC, Czech Republic
COST ES1206/GNSS4SWEC 3rd Workshop
March 8-10, 2016, Reykjavik, Iceland 1

2. Outline
Our new approach for NRT estimates using Kalman filter + backward smoother
Using GLONASS in the Benchmark campaign

3. Principle of the forward-backward filter
Filtering: Xk|k = E(X|y1,y2,y3, …, yk)
Smoothing : Xk|N = E(X|y1,y2,y3, …, yk, …, yN) 1 k N
y1, y2, y3, … yk
y1, y2, y3, … yk, … yN
Forward pass (Kalman, SRCF, SRIF)
Backward pass

4. Principle of the forward-backward filter
Flow chart of the algorithm

5. Principle of the forward-backward filter
Smoothing always improves precision w.r.t. filtering
Almost the same precision in all epochs
Observations from whole period contribute to all estimates - similar to LSQ, however with support for high-resolution estimates

6. RT/NRT estimates Kalman filter + backward smoother
Reference ZTDs: GOP’s Bernese (1h, network solution) and GFZ’s (15min, PPP)
Simulated RT ZTDs (IGS03 from IGS RTS): GOP’s G-Nut/Tefnut software (30 s, PPP)
NRT simulation (Kalman+smoother): different smoothing update: 15min, 1h, 2h, 4h, 24h

7. RT/NRT estimates Kalman filter + backward smoother

8. RT/NRT estimates Kalman filter + backward smoother
The precision of high-resolution ZTDs has been improved by smoother with delays up to 6 hours (up to 35% - from 10 mm to 6,5 mm)
Backward filter optimizes an estimate if any parameter changes very rapidly – it uses not only previous observations but also subsequent
Based only on matrix operation => very fast
(e.g. forward: 21 s; backward 2 s)
If real-time estimates are not necessary delayed results can be provided with better precision
Smoothed parameters over the whole period have been estimated with the same precision

9. Using GLONASS in the Benchmark campaign
How to optimally set up GLO observation weighting
Five solutions
Sol 1: GPS only: σGPS
Sol 2: GPS+GLO: σGLO = 2 σGPS
Sol 3: GPS+GLO: σGLO = 3 σGPS
Sol 4: GPS+GLO: σGLO = 4 σGPS
Sol 5: GPS+GLO: σGLO = 5 σGPS
RT: PPP, forward filter, 5min, IGS03 RTS
PP: PPP, forward filter, 5min, ESA final
Reference ZTDs: GOP’s Bernese (1h, network solution)
First two hours of a day are not included in the evaluation => no convergence

10. Using GLONASS in the Benchmark campaign
Sol 1
GPS
Sol 2
GPS + GLO
σGLO = 2 σGPS
Sol 3
σGLO = 3 σGPS
Sol 4
σGLO = 4 σGPS
Sol 5
σGLO = 5 σGPS

11. Impact of GLONASS on tropo gradients
GPS + GLO
σGLO = 2 σGPS
Products:
ESA final

12. Impact of GLONASS on tropo gradients
GPS + GLO
σGLO = 3 σGPS
Products:
IGS03 RTS

13. Conclusion
‘Ultra-fast’ troposphere estimates should be balanced for the delay & accuracy targeting the application
PPP with bi-directional filter are powerful approach for estimation large dense network in high-resolution
PPP provides full exploiting of multi-GNSS data
Our approach for NRT can be immediately applied in RT-Demo campaign - traditional hourly estimates with LSQ can be replaced by bi-directional filter with one hour delay
GLONASS improved troposphere estimates only when final products were used => low quality of RT products. (The same analysis should be done in for application in RT-Demo)

14. Thank you for your attention
Acknowledgements:
IGS for data and products - RTS, MGEX, Final
Benchmark GNSS data from EPOSA, SAPOS, ASG-EUPOS, CZEPOS, VESOG, GEONAS, Trimble
The work has been supported by the Czech Ministry of Education, Youth and Science (LD14102)
Related publications:
Václavovic P, Douša J (2015), Backward smoothing for precise GNSS applications, Advances in Space Research, Volume 56, Issue 8, 15 October 2015, Pages
Douša J, Václavovic P (2014) Real-time zenith tropospheric delays in support of numerical weather prediction applications. Advances in Space Research (2014), Vol 53, No 9, pp
Douša J, Dick G, Kačmařík M, Brožková R, Zus F, Brenot H, Stoycheva A, Möller G, Kaplon J (2016), Benchmark campaign and case study episode in Central Europe for development and assessment of advanced GNSS tropospheric models and products, Atmos. Meas. Tech. Discuss., discussion paper, 2016
Ahmed F, Václavovic P, Teferle FN, Douša J, Bingley R, Laurichesse D (2015), Comparative analysis of real- time precise point poisitioning zenith total delay estimates, GPS solut, FirstOnline, pp. 1-13